ai-paper-reproduction

Use when

• The user wants the agent to reproduce an AI paper repository.
• The target is a code repository with a README, scripts, configs, or documented commands.
• The goal is a minimal trustworthy run, not unlimited experimentation.
• The user needs standardized outputs that another human or model can audit quickly.
• The task spans more than one stage, such as intake plus setup, or setup plus execution plus reporting.

Do not use when

• The task is a general literature review or paper summary.
• The task is to design a new model, benchmark suite, or training pipeline from scratch.
• The repository is not centered on AI or does not expose a documented reproduction path.
• The user primarily wants a deep code refactor rather than README-first reproduction.
• The user is explicitly asking for only one narrow phase that a sub-skill already covers cleanly.
• The user is explicitly authorizing exploratory branch-only experimentation instead of trusted reproduction.

Success criteria

• README is treated as the primary source of reproduction intent.
• A minimum trustworthy target is selected and justified.
• Documented inference is preferred over evaluation, and evaluation is preferred over training.
• Any repo edits remain conservative, explicit, and auditable.
• Assumptions, protocol deviations, and human decision points are surfaced rather than hidden.
• repro_outputs/ is generated with consistent structure and stable machine-readable fields.
• Final user-facing explanation is short and follows the user's language when practical.

Interaction and usability policy

• Keep the workflow simple enough for a new user to understand quickly.
• Prefer short, concrete plans over exhaustive research.
• Expose commands, assumptions, blockers, and evidence.
• Avoid turning the skill into an opaque automation layer.
• Preserve a low learning cost for both humans and downstream agents.

Language policy

• Human-readable Markdown outputs should follow the user's language when it is clear.
• If the user's language is unclear, default to concise English.
• Machine-readable fields, filenames, keys, and enum values stay in stable English.
• Paths, package names, CLI commands, config keys, and code identifiers remain unchanged.

See references/language-policy.md.

Reproduction policy

Core priority order:

1. documented inference
2. documented evaluation
3. documented training startup or partial verification
4. full training only when the user explicitly asks later

Rules:

• README-first: use repository files to clarify, not casually override, the README.
• Aim for minimal trustworthy reproduction rather than maximum task coverage.
• Treat smoke tests, startup verification, and early-step checks as valid training evidence when full training is not appropriate.
• In trusted reproduction, a documented training command should first be checked through startup verification or a short monitoring window, then paused for explicit human confirmation before broader training continues.
• In explicitly authorized explore-lane execution, the training record can continue without the trusted-lane confirmation pause, but it must stay isolated from trusted conclusions.
• Record unresolved gaps rather than fabricating confidence.

Patch policy

• Prefer no code changes.
• Prefer safer adjustments first:
  • command-line arguments
  • environment variables
  • path fixes
  • dependency version fixes
  • dependency file fixes such as requirements.txt or environment.yml
• Avoid changing:
  • model architecture
  • core inference semantics
  • core training logic
  • loss functions
  • experiment meaning
• If repository files must change:
  • create a patch branch first using repro/YYYY-MM-DD-short-task
  • apply low-risk changes before medium-risk changes
  • avoid high-risk changes by default
  • commit only verified groups of changes
  • keep verified patch commits sparse, usually 0-2
  • use commit messages in the form repro: <scope> for documented <command>

See references/patch-policy.md.

Research safety boundary

• Preserve experiment meaning over convenience.
• Do not silently change dataset, split, checkpoint, preprocessing, metric, loss, or model semantics.
• Distinguish direct evidence from inference and from user-approved decisions.
• Prefer a recorded blocker over an unrecorded workaround.
• Escalate for explicit human review before any change that could alter scientific meaning or reported conclusions.

See references/research-safety-principles.md.

Workflow

1. Read README and repo signals.
2. Call repo-intake-and-plan to scan the repository and extract documented commands.
3. Select the smallest trustworthy reproduction target.
4. Call env-and-assets-bootstrap to prepare environment assumptions and asset paths.
5. Call analyze-project only when repo structure, insertion points, or suspicious implementation patterns need a read-only pass before continuing.
6. Run a conservative smoke check or documented inference or evaluation command with minimal-run-and-audit.
7. If the selected trustworthy target is documented training startup, short-run verification, or resume, hand execution to run-train instead of minimal-run-and-audit.
8. When training is selected inside trusted reproduction, let run-train capture the startup evidence first, then surface a human review checkpoint before any fuller training claim.
9. Stop for human review if protocol meaning, model semantics, or result interpretation would otherwise be changed implicitly.
10. Use paper-context-resolver only if README and repo files leave a narrow reproduction-critical gap that blocks the current target.
11. Never auto-route into explore-code or explore-run; exploration requires explicit user authorization.
12. Write the standardized outputs with evidence, assumptions, deviations, and next safe action.
13. Give the user a short final note in the user's language.

Required outputs

Always target:

repro_outputs/
  SUMMARY.md
  COMMANDS.md
  LOG.md
  status.json
  PATCHES.md   # only if patches were applied

Use the templates under assets/ and the field rules in references/output-spec.md.

Reporting policy

• Put the shortest high-value summary in SUMMARY.md.
• Put copyable commands in COMMANDS.md.
• Put process evidence, assumptions, failures, and decisions in LOG.md.
• Put durable machine-readable state in status.json.
• Put branch, commit, validation, and README-fidelity impact in PATCHES.md when needed.
• Distinguish verified facts from inferred guesses.

Maintainability notes

• Keep this skill narrow: README-first AI repo reproduction only.
• Push specialized logic into sub-skills or helper scripts.
• Prefer stable templates and simple schemas over ad hoc prose.
• Keep machine-readable outputs backward compatible when possible.
• Add new evidence sources only when they improve auditability without raising learning cost.
• Treat repo-intake-and-plan and paper-context-resolver as narrow helpers, not primary public entrypoints.